Reality Factory

I know a good poly count for RF is around 500-600, but whats the highest poly actor you can put into here?

dunno actually, don't try 16000, I tried once and the game threw me out, I was surprised (no I'm kidding, I wasn't ). My character models in my game are all above 1000 polys (which is bad, but at least the game looks good).
More than 1000 for me is not good!!! Here, on the picture bellow, the Police Battleship model is a little bit more than 3000 and take a look at the framerate (the ship s a static mesh):

And don't forget, Actor LOD is your friend

Ok. THanks for the input, but I'm asking for...the most it can handle without crashing.

Kiji, that depends on the detail of your scene. The more detailed your scene is, the less polys your actor can be. I would suggest trying to stay below 2500 if you can. This is also what normal maps are for. The correct way to use normal maps is to create a very detailed mesh, create your normal map from that mesh, then simplify it to the engine's specifications. Your original mesh could be 16000 polys and your simplified mesh w/normal mapping can be 2500 and you would not notice a difference (if done correctly).

You can throw in actors with more polys than 1000,if you have LODs for them.Even actors with base models of 3000 each.You just wont place many of them in the same place.
I d say you can even have 30 on screen and it is still playable.It gets a good frame hit if the 30 pawns have stencil shadows-this is tough to handle i d say without hardware acceleration.
Rf lacks (untill now) the feature to set different LOD distance for each actor.
The bigger the model,the bigger lod distances it must use.

Example:Small stones reach LOD3 -or LOD4 distance while a large stone would be in LOD1 distance values.
The higher the polycount the larger the space you must leave between them so LOD's can render in between.


About the high res model for creating normal maps,the polycount depends of your computer.The ussuall is 800.000 faces and the detail is projected on the low poly mesh.
Be sure the high poly model will give a texture that matches the UVW mapping coordinates of the low poly uvw's,because sometimes the tesselation of the high poly model distorts the UVW.

This is no problem,-you first render a smaller test texture to see if it matches,and if not,you can use projection mapping-a feature to project the mapping coordinates between the two models (high-low poly).
If you re going to displace the high res model,you ll have to edit the uvw after that to fix some distortion.

The high res model has so many vertices than it is difficult to handle (even with a fast computer) the UVW.
So the best way to do this,is to displace the mesh without tesselation and fix the stretching of the uvw.
What i do and works for me,is that i copy/paste the UVW modifier from the low poly to the high poly before the displace,so after the deformation the vertices match the UVW i pasted.

This is also what normal maps are for. The correct way to use normal maps is to create a very detailed mesh, create your normal map from that mesh, then simplify it to the engine's specifications. Your original mesh could be 16000 polys and your simplified mesh w/normal mapping can be 2500 and you would not notice a difference (if done correctly).

I has the grim feeling that normal maps don't do shading properly with RF, aqlthough possibly they work ok if its a static mesh?

The issue with normal maps in RF is that it works from a preset light direction only. It's not a hard fix. If the actor is rendered using DOT3, then grab the nearest 2 lights and use them to calculate the tangents.

Hmm so does that mean that if you had say a light that is animated to move around the actor, the normal map will have incorrect shadows?Or are you saying that when you first create the actor you have to have a light or 2 in the model file?Or am I just totally lost....

I am saying that the engine needs to know about the lights that are close to the actor in order to generate the tangent space vectors to use with the normal maps. If the lights move, the normal map should change with the direction of the light. That is the correct way to do normal mapping.

You cannot put lights into an actor file.

So if I create an actor with a normal map, and put it in RF, it will work?What more do I have to do to get it to work correctly?

If you want dot3 bumpmapping lightning direction to look steady,i say use it on terrain with only one spotlight above (with a huge radius).
This way everything on the terrain will have similar lightning-ground,stones tree barks etc.

I know the dark side will be black but,for the moment you have to use it as it is.
Even if two lights instead of one worked for the dot3,i think you would nt want to use a second dynamic (backlight) to an outdoor terrain map.Outdoor has one light source (except if it is some alien planet).

In the case where one only sky) light is used,you would need a higher ambient for the dot3 textured actors-which is not available yet -than the current 0 0 0 color.

Would normal maps look any good on player models?